Which Sume media tools take a 3-minute Reel and a 5-minute TikTok

Every Sume media tool except video captions accepts a 3-minute Reel (180 seconds) and a 5-minute TikTok (300 seconds); video captions is priced for videos up to 60 seconds, and video filter and frames stop at 300. Instagram Reels now allow 3 minutes per Metricool (read 2026-10-03), and TikTok is reported to favour 3 to 5 minute videos per Metricool (read 2026-10-03). The table gives the limits from the docs for each tool.

The numbers come from each tool's page in the docs; they are limits on the source file, not on what a platform accepts.

The limits at a glance

Check the right-hand columns: 180 and 300 seconds are the two lengths that matter this week.

How to read the table

A source limit bounds the file you hand in. For trim and detach there is also an output cap of 900 seconds, which only matters for long cuts. Timeline is bounded by the audio bed: audio.duration_seconds must be 1 to 1800, with up to 200 video slots and 20 audio parts.

Captions are the odd one: the fixed $0.20 estimate covers videos up to 60 seconds and the docs state no price above that. This post therefore prices a 3-minute Reel as three caption jobs and a 5-minute video as five. The trim to cut those parts costs $0.02 each, so the captioning bill for a 5-minute video is $1.00 plus $0.10 for the five trims, and a join render adds $0.50.

import json

CAPS = {  # source seconds from the Sume docs pages
    "video-inspect": 1800, "video-trim": 1800, "audio-detach": 1800,
    "video-filter": 300, "video-frames": 300, "video-captions": 60,
}

def fits(job, seconds):
    return seconds <= CAPS[job]

for minutes in (3, 5):
    s = minutes * 60
    print(minutes, "min:", json.dumps({j: fits(j, s) for j in CAPS}))

Worked examples

A 3-minute Reel: probe with inspect ($0.03 for 3 audio minutes), no cut needed, filter once ($0.02), and caption in three 60-second parts ($0.60 plus three trims at $0.06). Joined with a 3-minute Timeline render ($0.30) the total is about $1.01 before inspect's compute charge.

A 5-minute TikTok at the limit: inspect $0.05, filter $0.02 if the file is at most 300 seconds, caption in five parts $1.00 with $0.10 for trims, join $0.50. The sum is about $1.67 before inspect's compute charge. If the file is 301 seconds, filter will refuse it with output_duration_exceeded, so trim to 300 first at $0.02.

And a 20-minute webinar you want to turn into one 3-minute Reel: inspect once on the full source ($0.20, within 1800 seconds), trim the kept ranges, and render 3 minutes ($0.30). Nothing in that path needs a split because only trim and inspect see the long file.

What Sume does and does not do here

• Sume does not split long sources for you in any of these tools.
• Sume does publish the limits on each tool's docs page, and most jobs refuse with a stable error code rather than truncating (trim clamps an out-point past the source and warns).
• Sume does not post to Instagram or TikTok; it produces the files.
• Sume does not know a platform's current limit. The platform side of this page is reported by Metricool and may change.

What to do

Measure first. A single inspect call tells you the length and the audio. Then send each file to the tool whose limit it fits, and cut the file first when it does not.

Keep the table next to your pipeline code as a constant, as in the snippet above, and check each file against it before you submit. A refusal costs you a round trip; a local check costs nothing. Re-read the docs page of each tool when you update the constants, since limits are documented there and can change. For a mixed batch, group files by length first: files under 60 seconds can go straight to captions, files up to 300 seconds can use filter and frames, and anything longer goes through trim, then back into the right group. That grouping step is cheap and keeps the batch from failing halfway through, which is the real cost of guessing wrong on a long queue.

If you only remember one rule, remember the captions limit. Every other tool in the table takes the 5-minute file as is, so the plan for a long video is almost always to caption in 60-second parts and let the other jobs run on the whole.